AUTO-ENCODING VARIATIONAL BAYES

By Romain Lopez, Pierre Boyeau, Nir Yosef, Michael I. Jordan, Jeffrey C. Regier. 2020

Subfields: Generative Models, Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Generative Models

In Probabilistic & Bayesian Methods

Citations in the corpus

A Simple Framework for Contrastive Learning of Visual Representations (2020, neutral): “Generative approaches learn to generate or otherwise model pixels in the input space (Hinton et al., 2006; Kingma & Welling, 2013; Goodfellow et al., 2014).”
CTRL: A Conditional Transformer Language Model for Controllable Generation (2019, methodology): “Much research then focused on methods for controlling the generation process and improving estimation of generative distributions (Arjovsky et al., 2017; Chen et al., 2016; Kingma & Welling, 2013).”
DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning (2021, neutral): “In contrast, the approaches described above assume stochastic selection and are suitable in applications where there is an underlying sampling distribution, such as in variational inference [19].”
Are GANs Created Equal? A Large-Scale Study (2017, methodology): “Finally, we included the Variational Autoencoder [14] in the experiments as a popular alternative.”
Evaluating Lossy Compression Rates of Deep Generative Models (2020, methodology): “We use our rate distortion approximations to study a variety of variational autoencoders (VAEs) (Kingma & Welling, 2013), GANs and adversarial autoencoders (AAE) (Makhzani et al., 2015), and arrive at a number of insights not obtainable from log-likelihoods alone.”
VQ-DRAW: A Sequential Discrete VAE (2020, methodology): “This paper introduces VQ-DRAW, a new kind of discrete variational auto-encoder (VAE) [2].”
Predictive Sampling with Forecasting Autoregressive Models (2020, methodology): “First, an ARM is trained on images, we refer to this task as explicit likelihood modeling.”
BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling (2019, background): “One of the key aspirations in recent machine learning research is to build models that understand the world [24, 40, 11, 57].”
Neural Ordinary Differential Equations (2018, positive): “We can train this latent-variable model as a variational autoen-coder (Kingma and Welling, 2014; Rezende et al., 2014), with sequence-valued observations.”
Autoregressive Quantile Networks for Generative Modeling (2018, background): “…a staggering increase in progress on generative modeling in recent years, built largely upon fundamental advances such as generative adversarial networks (Goodfel-low et al., 2014), variational inference (Kingma & Welling, 2013), and autoregressive density estimation (van den Oord et al., 2016c).”
Generative Adversarial Networks (2021, neutral): “Since their first introduction in 2014, GANs have attracted a growing interest all over the academia and industry, thanks to many advantages over other generative models (mainly Variational Auto-encoders (VAEs) [34]):”
Projected GANs Converge Faster (2021, background): “Generally, GANs produce high-quality samples while being prone to mode collapse (high precision, low recall), compared to VAEs [39], which generate lower quality but more diverse samples (low precision, high recall).”
Vector-quantized Image Modeling with Improved VQGAN (2021, background): “Many image generation models (Goodfellow et al., 2014; Kingma & Welling, 2014; Radford et al., 2016; Donahue et al., 2017; Higgins et al., 2017) have been studied for their capabilities in representation learning.”
Dual Contradistinctive Generative Autoencoder (2020, neutral): “The objectives of a variational autoencoder (VAE) [37] is to simultaneously train an inference network q φ ( z | x ) and a generator network p θ ( x | z ) .”
High Fidelity Visualization of What Your Self-Supervised Representation Knows About (2021, neutral): “…latent variable models and variants of auto-encoders (Ack-ley et al., 1985; Hinton et al., 2006; Salakhutdinov et al., 2007; Vincent et al., 2008; Kingma & Welling, 2014; Rezende et al., 2014), that are traditionally put under the broad umbrella term of unsupervised learning (Bengio et al.,…”
On Mutual Information Maximization for Representation Learning (2019, background): “A similar phenomenon was observed in the context of variational autoencoders (VAEs) (Kingma and Welling, 2014), where one maximizes a lower bound on the data likelihood: Looser bounds often yield better inference models, i.e. latent representations (Rainforth et al., 2018).”
Towards Interpretable Natural Language Understanding with Explanations as Latent Variables (2020, methodology): “Due to the large search space of natural language explanation e, instead of using reprametrization trick used in variational autoencoder [14], here we use the variational-EM algorithm for optimization.”
High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs (2017, methodology): “We experimented with the Kullback-Leibler loss [27] on the feature space for better test-time sampling as used in the recent work [64], but found it quite hard for users to directly adjust the latent vectors for each object.”